TDS540A Acquisition Board problem

[mwbarth36]

  Hi! Can anyone help?

I really need help with a problem I have with a TDS540A acquisition board (A10).  Actually, I have two boards with a similar problem.  I have, in the past several years, fixed several of these scopes and am familiar with the electrolytic capacitor problem.  All electrolytic capacitors have been replaced, and the board has been washed ( a couple of times).  I did not find any open connections, the leakage did not appear bad enough to cause any open.  First let me describe the problem.

On startup, the to startup tests, 'ACQUISITION' and 'ATTN/Acq Interface', FAIL.  The attached photo shows the display with a square wave input.  Channel 2 displays the square wave exactly as it should, while Channel 1 appears to display the waveform, but with huge spikes.  Channels 3 and 4 are shown in the second photo.  Again the signal is present but with the voltage spikes present.  With no signal, the spikes are still present on the base line.  They do appear to change amplitude as the vertical position control is changed.

I know it should be obvious with this information where to start looking, but it just isn't!  If all channels behaved the same, I would look reference signals all channels used, and if only one channel were bad, I would look at that channels circuits.  With three bad channels, just not sure where to start looking.  What's more, I cannot find any signals from the three bad channels that are missing, or differ from the signals on the good channel. 

Has anyone seen a similar problem, or do you have any suggestions....................Please HELP!

[shakalnokturn]

If you can't trace anything different on the analog (FE to ADC) side of things would it make sense that the problem could come from the digital side of the acquisition?
The most likely to me would be one or more broken lines between the MUXes and acq. RAM, or one of these components being defective.
Sampling a waveform with 2 and 4 channel interleave could give you some extra clues.

[mwbarth36]

Thank you for your reply.  Your take on the problem makes sense, especially since the signal appears to be there on the three bad channels.  Wonder how Channel 2 can escape the problem?  I will do some more looking into some of the signals dealing with getting the data to the appropriate memory.  Thanks again, any further ideas on how to isolate this problem will be appreciated.

[Dacke]

I restored the TDS540 that I'm still using now,  also had to completely recap it.  Even after the recap it still had issues however,  until I identified some damaged traces that I had missed near some opamps that had lost their power rails.  You really have to go over the acquisition board very closely and even ohm out traces near wherever the caps had leaked...  Also make sure to clear the memory periodically and manually run the diagnostic tests and SPC - when I had finally repaired everything it still would not pass the startup diagnostics until I cleared everything out and ran them myself.

[shakalnokturn]

It's true that the section of the PCB that contains the analog MUXes and buffer OP-AMPs that share the FE control voltages out from the DAC is usually severely corroded by the capacitor leak.
I doubt that a defective OP-AMP could cause those symptoms, a bad MUX (or not switched due to cut track) could though. Try checking the FE offset control voltages, you should have DC, if not there's something wrong.

[Dacke]

It's been a couple years since I worked on it but yes my symptoms were different than his,  I believe the symptoms were inability to trigger/lock on to a signal, "gaps" in the sampling (the trace would draw a sine wave but with a dotted line), and unstable relays in the attenuator hybrids - relays for switching the input impedance and attenuation would randomly turn on and off regardless of input.  The first two issues were damaged traces near opamps,  the third issue was caused by damaged relay driver ICs on the attenuator board.

[feedback.loop]

What are the error messages in the log?

[mwbarth36]

I have attached the Error Report just run, only the last two are from the latest run.  I do not know why there are so few errors reported, they must get cleared at some point, because I have restarted this scope many times over the past month. 

Also have checked the reference voltages at the input to the DACs, all are DC levels consistent with the +/- levels one would expect.  However, the outputs from the DACs of the bad channels, 1,3, and 4 are all very different from the outputs from the DAC of the good channel, 2; even with the same input and vertical setting for all channels.  It makes me believe it is a DAC problem, but all inputs look normal, and would all three DACs come up with the same problem? 

One thing I noticed, the outputs from all four preamps go to all of the DAC blocks.  Do not know how they are processed in that block, but it would appear that something wrong with any channel preamp out (like an open connection) would, could, cause a problem on any, or all, channels.  I am going to investigate that inter-relationship of the preamp output next.  Thanks for all of the suggestions!

[feedback.loop]

The log seems to suggest a faulty RAM somewhere. I think I have seen a mapping between addresses and channels in one of repair threads on this site. Not sure you can narrow down to a particular chip especially with log lines cut off like this. Do you happen to have access to a GPIB interface adapter? It is possible to read full lines through GPIB. I am not sure it will help much though without detailed mapping from addresses and bits to chips.

[shakalnokturn]

One thing I noticed, the outputs from all four preamps go to all of the DAC blocks.

That's for the single and dual channel use, they interleave 2 (dual channel) or all 4 (single channel) ADC's for higher sample rate. Each ADC samples at a maximum of 250MHz but they take it in turns to give an equivalent of 1 GS/s.

The reason I suggested testing in 1 and 2 channel mode previously.

[mwbarth36]

An update on what I found over the week end.  The error log does appear to indicate a memory problem, but the information in the log is not very useful.  I do not have access to a GPIB board at this time.  Although this memory problem could be part of the problem, there appears to be another problem with the operation of the ADC's that is causing the very strange displays.  I say this because the outputs of the ADC for channel 2 (the working channel) is very different from the other three cannels (the bad channels).  I do believe it has to do with the interleaving of the information in the ADC's.  I am not at all sure how this works, take a look at how the Preamps and the ADC's are wired (see the wiring diagram in the attached drawing). 

At this point, when all four channels are displaying a common waveform input all preamplifier outputs have analog signal (which look normal, about 60mV for the signal I am using) except for +/- Sig34CD.  Not sure if that is normal or not? Output of the ADC's, however, is very different.  The digital signal on the bad channels is locked at the rails for long periods of time (which would account for the large spikes on the display) while the good channel is more what one would expect when digitizing a slow analog sinewave.

Any comments appreciated!! 

[feedback.loop]

Have you tried running signal path compensation?

[shakalnokturn]

What your drawing shows is that (unlike some other 4CH DSO's of the time) the TDS allows you to do 2 or 4 channel interleave using any input (1,2,3,4 identify the inputs. A,B,C,D the ADC's).

I expect something is wrong if 34CD outputs differ from the others, in non-interleaved sampling each input should have it's own ADC (1->A, 2->B....).

I'd beware of the error log messages pointing to RAM without knowing exactly how the system goes about testing RAM. If data is written directly from CPU board to Acq RAM then read back checking consistency it can be trusted if instead it is injected through the FE by (i.e.) hitting min and max limits on the ADC's it could be misleading. (I don't know how it's done, someone like madao maybe does?)

At this point I'd stick to what you've actually seen: One suspicious differential analog output and at least two suspicious DAC outputs.
It isn't impossible that components are actually damaged considering someone may have given it a try before and wrecked things all the same I'd look towards the sampling timing and control source, possibly compare ADC clocking, reference voltages... between working and defective channels.

[mwbarth36]

I had not thought to do SPC with the problems this scope showed, but did so knowing it would fail.  It failed (of course) but did provide some more information in the error log, see the attachment.  I then ran just the Acquisition test, and the the screen showed:  Acquisition/ Attn PASSED, Acquisition FAILED. all useful information. 

The other surprising fact that I did not expect is that the Sig34CD that was missing is now back and looks as it should.  In fact, signals at the the Preamp outputs all look normal now, I am convinced the problem is not with the Preamp section.  Will look more at the ADC's tomorrow. 

[mwbarth36]

I have some additional information in case anyone is still following this thread.  I believed I can safely say that the problem is not with the ADC's or the Demultiplexers.  First, I connected the same triangular input to all four channels, and enabled only channel one (a bad channel).  Photo one shows the output A7 of U850, the ADC for channel one.  In photo one the bottom trace is the input signal, the top trace is A7 output of U850.  As can be seen, bit A7 changes exactly at the zero crossing of the +/- 2 Volt triangular wave input.  When each channel is enabled individually, the corresponding A7 bit or each ADC responds in a similar manner.  This, I believe is what one would expect since this bit responds to the polarity of the input signal.  Now observe the displayed waveform on the TDS540, photo 2.  The basic triangular waveform is present, but broken up by the seemingly random spikes.  Nothing in the output of the ADC for channel 1 would indicate that these spikes are being caused by anything forward from the ADC.  It is interesting that the output B7 of the Channel one ADC is identical to the bit A7 output.  This is true for all channels when enabled individually.  Not sure why the two outputs for each channel, I suspect it has something to do with the demultiplexing.

Speaking of demultiplexing, I also believe that the demultiplexers are working just fine, as most of the signal (all of the signal for channel 2) appears as it should for all channels (except for the spikes.  Since all channels are turned off except for channel one, I would expect that if the demultiplexers were messing up, any demultiplexer screw up would cause the displayed signal to have any spikes back to the zero level, which is clearly not the case.

This leaves the memory, or memory control, but this would mean that there is bad memory in three of the four channels, keep in mind that channel 2 works just fine.  Indeed, this seems to eliminate the control circuits.  It is just difficult to believe that three of the four channels would develop bad memory together. 

All comments are appreciated.  I am sure I am a victim of 'fuzzy thinking' here, but cannot determine where.  Any ideas on how to prove it is a memory problem?  Or not  a memory problem?

[mwbarth36]

No new ideas??  I am determined to understand this problem, but have not made any progress since last post.  Any ideas as to how to check the memory chips without simply replacing them all?  Still do not understand why there would be faulty memory chips in all channels.  Not a likely situation.

[madao]

Bad  SRAM or broken trace between  Multiplexer and SRAM.

A good notice from TDS520 Service manual is attached as picture.

[mwbarth36]

Thank you for the information.  While I have not yet located the problem with this scope, the existence of the 'Component Level Diagnostic and Repair Manual', part No. 070-8313-00 is a great help and should eventually lead to solving the problem.  This manual keeps referring to Diagnostic programs for specific problems which are not available using the scope's front panel controls.  I am assuming that these Diagnostics are accessible through the GPIB port.  Do you know if there is a document describing these tests, and just how to access them.  I do not currently have an interface for this port, but would like to get one if the tests are as comprehensive as they appear to be.    A manual describing how to access would be of great help.  Thanks again for the information.

[shakalnokturn]

Correct me if I'm wrong but I think mentioned manual is for the TDS520B while I expect it to be very close to the older TDS500's and very useful for general understanding, there are some differences in hardware such as ADC to data MUXes having gone differential on the B models.

Although I've never used it I know about the Tektronix Field Adjustment Software (FAS) it's available online if you look around, it requires MS-DOS and a (specific model of) lSA GPIB card and to interface with the TDS. I don't know if the diagnostics routines are part of it or from separate software though.

[james_s]

I fixed a similar problem with my TDS784C, I don't remember what the exact error message was but in my case it was intermittent. I fixed it by resoldering the big QFP ICs on the acquisition board. Be really careful if you attempt this and inspect very carefully for solder bridges before you apply power.

[mwbarth36]

The manual is actually for the 520.  The Acq board A10 is very different in the 520B.  So the good news is the CLDR Manual is for the scope I have and has been very useful in determining that the problem is not in the Preamp section or the A/D converter section.  When it comes to the Demux section it is almost impossible to trouble shoot without running the diagnostics.  I have seen the Tektronix Field Adjustment Software advertised, does anyone know if the diagnostic software is part of that?  Still, needing a DOS machine and special card to run it does not seem a practical way to approach things right now. 

I believe my next step is going to be as James_s suggests, resolder the Demux chips, they do not look like the best connections, and after so many years, who knows??

Thanks for all the help!

[mwbarth36]

An update on the 540 acquisition board problem.  I did re solder the demux chips for channels one and two.  That produced no change.  I then started looking at the data outputs A0-A7 and B0-B7 on each of the memory chips with only channel one active, channel 1.  On most of the memory chip inputs I observed clean zero to 5 volt signals, except for two of the chips.  Several of the data lines on these chips were not able to either go to zero volts, or +5 volts.  So, to see if it was a demux problem or a memory chip problem, I removed the suspected bad memory chip, and checked the previously bad lines again.  The signals from the demux, without the memory chip were as they should be, clean 0 to 5 volts signals.  Since each line from a demux only goes to one pin on the memory chip, the demux appears OK.  The memory chip must be the problem.  Another indication that the problem is with the memory chip was that removing the bad chip did not change the pattern produced on the screen.  I then checked all of the other channels, one at a time, and found bad signals in all of the bad channels, and none in the memory for channel 2, the channel that does not have the problem.  I do believe that I have pinned the problem to multiple failed memory chips.  Waiting for some replacement chips to see if I am correct.

[james_s]

That certainly sounds promising, nice job. For what it's worth, SRAM has been one of the more failure prone parts I have encountered over the years in various things.

[madao]

Yes,    SRAM fail is not rarely by this Units.  Broken trace is more common than failed SRAM.  IDT SRAM fails very often, if it is wetted with eletrolytics.
6264 SOP is not very common,  more common is  6264 SOJ
62256 SOJ is cheap. (i have buy 100 pieces with 20ns speed for $20 )

I have replace doze  SRAM by bigger 62256 and solder it with brigde between Pin 1 and 2 and  Pin 2 ist bent as  SOP-lead.
Then place 62256 on 6264s footprint (17-32 pin) and Pin 1-14 of 6116-footprint.  modified 62256 pin 2 lead is connect to  62256 footprint pin 2.

Pic show all failed SRAM from ONE TDS540

[max-bit]

There are multi-layer boards used, I don't remember exactly how many, but maybe 6 or even 8 layers
The fact that you do not see the damage in  does not mean that there is no damage, because probably some grommet has been damaged. Checking this out is not so much difficult as it is impossible.
Tektronix (as it was still in support) did so that simply replaced the entire PCB, and in fact, I would also propose to buy a new (read) working PCB.